The principle of wavelength division multiplexing consists of transporting, on a single optical waveguide such as a fiber, a plurality of independent data signals which respectively modulate a plurality of optical carriers occupying distinct optical wavelengths. This allows for significant savings when it is desired to increase the capacity of a network that already has optical fiber in place but where the fiber was previously used for transporting only a single optical carrier occupying a single optical wavelength. Since an optical carrier is implicitly associated with an optical wavelength, the expressions “optical carrier” and “optical wavelength” will hereinafter be used interchangeably.
In a wavelength division multiplexed (WDM) network, each optical carrier is associated with its own source and destination nodes. Where multiple optical carriers have intersecting routes, these multiple optical carriers will occupy different wavelengths of light on the same fiber. When this type of multi-carrier signal travels along a long route, amplifiers will be required at every 80 kilometers or so in order to boost the signal's optical power.
On even longer routes, a multi-carrier optical signal may not just suffer severe attenuation but it may also become distorted due to effects such as chromatic dispersion, polarization mode dispersion, signal-to-noise ratio degradation resulting from noise contributions of multiple cascaded amplifiers, and non-linearities in the optical transmission medium or in the optical components traversed along the way.
Distortion of this nature is sometimes counteracted by inserting equipment in the optical path for providing dispersion compensation or banded gain equalization. In severe cases of distortion, an array of regenerators may need to be added. In its most basic form, a regenerator array detects the data on each incoming carrier and uses the detected digital data to re-modulate a fresh (usually re-shaped and re-synchronized) optical signal on the appropriate optical wavelength. Thus, a regenerator array requires, for each wavelength it is required to regenerate, an optical receiver, electronic re-shaping and re-timing circuitry and an optical source. For a dense wavelength division multiplexing (DWDM) system with typically 32 to 160 wavelengths per fiber, this leads to a very complex regenerator array.
In order to allow the flexible interconnection of optical carriers, an optical network must be equipped with a mechanism for providing switching functionality at the optical carrier level. Traditionally, an optical interconnect mechanism is implemented either as an optical patch panel or as an electrical switch (or cross-connect) with optical-to-electrical converters at its input and electrical-to-optical converters at its output.
A cross-connect differs from an switch in that for the case of a cross-connect, the connection map is usually provisioned from a central network management tool, either automatically or manually, whereas for the case of an optical switch, the connection map can be controlled in real time and may even be controlled by the traffic content through the switch, in which case the switch is said to be self-routing. In the interest of simplicity, and because a switch inherently encompasses a cross-connect as well as a switch in the strict sense of the term, references made to a switch in the remainder of the specification should be understood to mean a cross-connect or a switch, depending on the circumstances.
While an electrical switch provides adequate switching functionality for a low density of wavelength division multiplexing, i.e., to a small number of optical carriers per fiber, those skilled in the art will appreciate that as the density of a WDM optical network grows, it becomes prohibitively expensive (both pecuniarily and from the point of view of power consumption) to equip an electrical switch with sufficient optical-to-electrical and electrical-to-optical conversion resources to handle multiple incoming dense WDM signals arriving on their respective optical fibers.
To this end, the art has seen the development of the “photonic” switch (or cross-connect), which is the counterpart to the electrical switch (or cross-connect). In a photonic switch, switching is performed almost purely in the optical domain with only minimal recourse to optical-to-electrical or electrical-to-optical conversion. This advantageously results in significant reductions to the cost and complexity of the switching equipment.
A photonic switch can take on many generic forms, one of which is shown in FIG. 1 and more fully described in above-referenced co-pending U.S. patent application Ser. No. 09/511,065. The photonic switch 100 typically comprises N individual M-output wavelength division demultiplexing (WDD) devices 110A-110N, where each WDD device is associated with a respective one of N input fibers 120 connected to a respective set of N amplifiers 125. The photonic switch 100 also comprises N individual M-input wavelength division multiplexing (WDM) devices 130A-130N, one WDM device for each of N output optical fibers 140 connected to a respective set of N amplifiers 145.
The photonic switch 100 also comprises a photonic switch core 150 connected between the WDD devices 110A-110N and the WDM devices 130A-130N and a switch controller 160 connected to the photonic switch core 150.
On the input side of the photonic switch 100, each of the N WDD devices 110A-110N accepts a respective input WDM signal on a respective one of the input optical fibers 120. Each of the N WDD devices 110A-110N then separates the respective input WDM signal on a per-wavelength basis into a plurality (M) of input individual optical carrier signals which are provided to an input side of the photonic switch core 150 along a respective plurality (M) of demuxed input optical paths 170, which may consist of optical fibers, silica waveguides or other optical transmission media.
The photonic switch core 150 switches the input individual optical carrier signals, thereby to produce a plurality of switched individual optical carrier signals which are carried out of the photonic switch core 150 by a plurality of demuxed switched optical paths 180. The switch controller 160 generates a connection map under external or locally generated stimulus, which connection map is provided to the photonic switch core 150 and defines the desired map of the optical channels from the input side to the output side of the photonic switch core 150. External stimulus may be provided via a control link 165.
At the output of the photonic switch core 150, each of the WDM devices 130A-130N receives a respective set of demuxed switched optical paths 180 and combines the switched individual optical carrier signals thereon into a single respective WDM signal that exits the photonic switch 100 along a respective one of the output optical fibers 140.
In the illustrated embodiment, the photonic switch core 150 comprises a wavelength converting switch 190 and M optical switch matrices 150A-150M, one for each of the M optical wavelengths in the system. Each optical switch matrix has a set number of input ports and output ports and can be a Micro-Electro-Mechanical System (MEMS) device as described in “Free-Space Micromachined Optical-Switching Technologies and Architectures” by Lih Y. Lin of AT&T Labs-Research during OFC99 Session W14-1 on Feb. 24, 1999. This article is incorporated by reference herein. Such a MEMS device comprises a set of mirrors that are arranged in geometrical relationship with the input and output ports such that incoming light from any input port can be diverted to any output port by erecting an appropriate one of the mirrors under control of the switch controller 160.
In FIG. 1, each of the optical switch matrices 150A-150M has a total of K+N input ports and K+N output ports where, it is recalled, N is the number of WDD devices 110A-110N and WDM devices 130A-130N. For each of the optical switch matrices 150A-150M, each of the N input ports will be connected to the like-wavelength output port of a respective one of the WDD devices 110A-110N, while each of the N output ports will be connected to the like-wavelength input port of a respective one of the WDM devices 130A-130N. This permits optical signals of a given wavelength entering a switch matrix 150A-150M to be connected to the appropriate port of any of the exiting WDM devices 130A-130N.
It is thus noted that each of the optical switch matrices 150A-150M has K more input ports and K more output ports than are required to switch the N corresponding input individual optical carrier signals (one of which arrives from each of the N WDD devices 110A-110N). These additional ports are connected to the wavelength converting switch 190, with two important consequences. Firstly, optical carrier signals arriving on demuxed input optical paths 170 can be redirected towards the wavelength converting switch 190. Secondly, optical carrier signals arriving from the wavelength converting switch 190 can be output onto one of the demuxed switched optical paths 180.
The net result is that a signal on an individual optical carrier is allowed to change wavelengths on its way through the photonic switch 100 by a process which involves optical reception, opto-electronic conversion, electrical switching of the converted electrical signal to an optical source at a desired wavelength and modulation of that source's optical output. The wavelength conversion process is particularly useful when an input wavelength is already in use along the fiber path leading to a destination WDM device.
It should further be noted that the wavelength converting switch 190 also accepts a plurality of “add carriers” on a plurality (R) of add paths 192 and outputs a plurality of “drop carriers” on a plurality (R) of drop paths 194. Thus, it is seen that the wavelength converting switch 190 has a total of ((K×M)+R) inputs and a like number of outputs. Structurally, the wavelength converting switch 190 comprises a set of ((K×M)+R) electrical-to-optical converters, an electrical switch and a set of ((K×M)+R) optical-to-electrical converters that collectively function as a miniature version of an electrical switch for optical signals.
The term “wavelength converting switch” will hereinafter be used throughout the following, with the understanding that such a switch may have either purely wavelength conversion capabilities or both wavelength conversion and add/drop capabilities.
In operation, the photonic switch 100 of FIG. 1 provides purely optical switching at the optical switch matrices 150A-150M and wavelength conversion (most commonly through the use of electrical switching) at the wavelength converting switch 190. Control of which input individual optical carrier signals are redirected into the wavelength converting switch 190 is provided by the switch controller 160. The switch controller 160 also provides control of the switching executed inside the wavelength converting switch 190.
With the assistance of network-level control of the wavelengths used by the various sources in the network, it is usually possible to ensure that most wavelengths can transit directly across most nodes in the network without wavelength conversion, hence ensuring that the majority of optical carriers will be sent along the desired output optical fiber 140 directly by the optical switch matrices 150A-150M without involving the wavelength converting switch 190. As a result, it is usually possible to achieve a minimal blocking probability at the photonic switch 100 by selecting a relatively small value for K, i.e., by keeping most of the switching entirely in the optical domain.
The photonic switch described in part herein above and described in more detail in co-pending U.S. patent application Ser. No. 09/511,065 is an example of how developments in the field of optical switching are often stimulated by the need to accommodate the ever increasing optical wavelength density of WDM networks in general and WDM signals in particular.
In addition, the increase in density has driven up the cost associated with providing optical signal regeneration. This is largely due to the higher number of optical sources and receivers which must be provided at a regenerator site in order to handle the increased number of optical carriers per fiber, since each optical carrier has to be regenerated separately and independently. Consequently, those skilled in the art have begun to concentrate on lowering the cost of regeneration by trying to expand the reach between optical regeneration points in a dense WDM network.
The reach between optical regeneration points is limited by the build-up of degradation suffered by the optical carriers in the WDM signal which cannot be removed (and may actually be introduced) by current optical amplifiers. Specifically, the maximum reach attainable between first and second regeneration points is limited by factors such as:                launch power and pulse shape at the first regeneration point;        receiver sensitivity at the second regeneration point;        build-up of uncompensated chromatic dispersion and polarization mode dispersion along the route;        accumulation of noise arising from cascades of intervening amplifiers;        excessive flat gain or loss of intervening amplifiers, WDM/WDD elements, connectors, splices and fibers;        wavelength-dependent gain or loss through intervening amplifiers, WDM/WDD elements, connectors, splices and fibers; and        cross-modulation and inter-modulation effects.        
Many of the above factors contribute to producing a non-flat optical power spectrum of the WDM signal, i.e., the individual optical carriers will experience different amounts of gain and loss as they propagate. The resulting WDM signal with a non-flat optical power spectrum will reduce the maximum reach because optical carriers having higher power may saturate the intervening optical amplifiers, while optical carriers having lower power may not be detected with sufficient accuracy by a far-end regenerator. Consequently, the power differential between high power carriers and low power carriers has to be minimize in order to maximize the reach between regenerators.
In attempting to solve this problem, it has been realized that for a conventional, point-to-point WDM system, variations in the optical power of the component carriers of a WDM signal are often correlated between one optical carrier and its neighbours in the optical spectrum, due to having undergone a common, wavelength-dependent amplitude distortion process. Conventional spectrum flattening techniques take advantage of this realization to provide “band equalization” of the power spectrum at an intermediate component between two regenerators. This type of equalization technique is now described with reference to FIG. 2.
Specifically, a band WDD device 4 may be used to separate an original WDM signal arriving on an input optical fiber 2 into a plurality of separate optical paths each consisting of a number of signals occupying mutually exclusive optical frequency bands. For simplicity of illustration, there are three groups of signals occupying three bands denoted A, B, C, but there may be five bands in a typical band equalization scenario. The three separated groups of signals are still WDM signals in their own right but have fewer carriers than the original WDM signal.
Each of the three signals in bands A, B, C passes through a respective one of a plurality of variable optical intensity controllers (VOICs) 6, 8, 10. Each of the VOICs 6, 8, 10 could be an amplifier or an attenuator having a response which is controllable within the band of interest but is irrelevant elsewhere. The outputs of the three VOICs 6, 8, 10 are then recombined by a band WDM device 12 into a recombined WDM signal provided on an output optical fiber 14.
In FIG. 2, the optical power spectrum of the original WDM signal on the input optical fiber 2 is shown at 16 and, in this example, is seen to comprise a total of fifteen optical carriers, five in each of the three broad optical frequency bands A, B, C. The correlation among the power levels of neighbouring carriers in the input optical power spectrum 16 is apparent from the diagram.
In addition, it is seen that the overall peak-to-peak power level variation (shown at 18) of the input optical power spectrum 16 is significant. However, because of the correlation among the power levels of neighbouring carriers, it is possible to identify an average power level 19A, 19B, 19C in each respective band such that the peak-to-peak power level variation with respect to that average power level in that band is reduced as compared to the overall peak-to-peak power level variation 18.
In order to achieve band equalization, the gain (or attenuation) to be applied by each of the VOICs 6, 8 and 10 is set to a value which complements the estimated average power level in the corresponding band in order to bring the average power level to a target level. Since the band equalization is usually a static technique, average power level estimates can be obtained at installation time. In the case of FIG. 2, comparing the average power levels 19A, 19B and 19C in bands A, B and C (which can be estimated at installation time), it is seen that VOIC 6 should be accorded a moderate gain, VOIC 8 should be accorded a high gain and VOIC 10 should be accorded a low gain.
After applying band equalization in the manner of FIG. 2, the optical power spectrum (shown at 20) of the recombined WDM signal provided on the output optical fiber 14 is seen to have a significantly lower overall peak-to-peak power level variation (shown at 22) when compared to the overall peak-to-peak variation 18 in the original WDM signal.
However, it will be apparent that the band equalization approach does not completely remove peak-to-peak variations in the optical power spectrum of the original WDM signal. Rather, it provides a mechanism for reducing the level of variation and results in this level of reduction being traded off against implementational complexity by exploiting the correlation existing between adjacent carriers. Therefore, as seen in FIG. 2, the resultant WDM signal travelling on the output optical fiber 14 still contains wavelength-dependent variations in its optical power spectrum 20.
Furthermore, the band equalization technique illustrated in FIG. 2 does not account for wavelength-dependent power level variations which may have been introduced by the band demultiplexer 4 and the band multiplexer 12. Although not explicitly shown in FIG. 2, the optical power spectrum 20 of the output WDM signal could conceivably contain even more significant variations due to the compounded effects of the band demultiplexer 4 and the band multiplexer 12.
A further cause of variance in the optical power spectrum of a WDM signal is the action of a photonic switch such as that shown in FIG. 1. Specifically, because the connection map of the photonic switch is arbitrary, being driven by traffic connectivity considerations rather than optical link considerations. Thus, a particular output WDM signal emerging from the photonic switch will contain optical carriers that will likely have travelled along entirely different paths through the network. Each of these paths is associated with its own loss characteristics and therefore the various individual optical carrier optical signals that make up a WDM signal at the output of the photonic switch will have respective optical power level which are uncorrelated with one another.
The situation is illustrated in FIG. 3, where a 3×3 photonic switch 300 is connected to three input optical fibers 40, 42, 44 and three output optical fibers 60, 62, 64. The input optical power spectrum of the WDM signal on each of the input optical fibers 40, 42, 44 is shown at 50, 52, 54, respectively. Each of these three input optical power spectra 50, 52, 54 occupies the same optical frequency range but has a distinct shape. In particular, the shape of each of the optical power spectra 50, 52, 54 displays a certain degree of correlation among the power levels of neighbouring carriers. For example, spectrum 50 has a monotonically decreasing shape, spectrum 52 has a bell shape and spectrum 54 is composed of relatively constant power levels.
Since any arbitrary connection map may be provided by the photonic switch 300 at a given instant in time, the correlations existing among the carrier power levels on a the input optical fibers 40, 42, 44 may not carry through to the output optical fibers 60, 62, 64. Hence, the output optical spectra (shown at 70, 72, 74) will exhibit a poor correlation among individual carriers and will appear “randomized”. This effect may be compounded by differing losses experienced by the various signals as they transit the switch node components. Clearly, as a result of this lack of correlation among individual carriers, a band equalization technique such as that previously described with reference to FIG. 2 would be of little use if applied at the output or even at the input of the photonic switch 300.
Those skilled in the art will also appreciate that in addition to being affected by spectral variations arising from the arbitrary connection map applied by a photonic switch, the optical power spectrum of an output WDM signal may be further distorted by wavelength-dependent losses induced by a WDM device positioned at the output of the switch and, to a certain extent, by path-dependent losses through the photonic switch core.
Hence, it will be appreciated that the optical power spectrum of the WDM signals exiting a photonic switch can be severely distorted and, worse still, the distortion has no predictable spectral shape. Moreover, the optical power spectrum of the WDM signals can change dramatically and suddenly with each change in the connection map. Clearly, such wavelength-dependent distortion presents a serious limitation on the reach between the photonic switch and the next regeneration point in the network and therefore it would be a tremendous advantage to provide spectral flattening at the photonic switch, without adding significant complexity to the design of the photonic switch itself.